Prediction of Daily Mean PM10 Concentrations Using Random Forest, CART Ensemble and Bagging Stacked by MARS

A novel framework for stacked regression based on machine learning was developed to predict the daily average concentrations of particulate matter (PM10), one of Bulgaria’s primary health concerns. The measurements of nine meteorological parameters were introduced as independent variables. The goal was to carefully study a limited number of initial predictors and extract stochastic information from them to build an extended set of data that allowed the creation of highly efficient predictive models. Four base models using random forest, CART ensemble and bagging, and their rotation variants, were built and evaluated. The heterogeneity of these base models was achieved by introducing five types of diversities, including a new simplified selective ensemble algorithm. The predictions from the four base models were then used as predictors in multivariate adaptive regression splines (MARS) models. All models were statistically tested using out-of-bag or with 5-fold and 10-fold cross-validation. In addition, a variable importance analysis was conducted. The proposed framework was used for short-term forecasting of out-of-sample data for seven days. It was shown that the stacked models outperformed all single base models. An index of agreement IA = 0.986 and a coefficient of determination of about 95% were achieved.

Application of GARCH Forecasting Method in Predicting The Number of Rail Passengers (Thousands of People) in Jabodetabek Region

PT. Kereta Api Indonesia and PT. KAI Commuter Jabodetabek records time series data in the form of the number of train passengers (thousand people) in Jabodetabek Region in 2011-2020. One of the time series methods that can be used to predict the number of train passengers (thousand people) in Jabodetabek area is ARIMA method. ARIMA or also known as Box-Jenkins time series analysis method is used for short-term forecasting and does not accommodate seasonal factors. If the assumption of residual homoscedasticity is violated, the ARCH / GARCH method can be used, which explicitly models changes in residual variety over time. This study aims to model and forecast the number of train passengers (thousand people) in Jabodetabek area in 2021. Based on data analysis and processing using ARIMA method, the best model is ARIMA (1,1,1) with an AIC value of 2,159.87 and with ARCH / GARCH method, the best model is GARCH (1,1) with an AIC value of 18.314. Forecasting results obtained based on the best model can be used as a reference for related parties in managing and providing public transportation facilities, especially trains.

Assessing Forecasting Performance of Daily Mean Temperature at 1st and 2nd Perak Station, Surabaya Using ARIMA and VARIMA Model with Outlier Detection

Air temperature is an important data for several sectors. The demand of fast, exact and accurate forecast on temperature data is getting extremely important since it is useful for planning of several important sectors. In order to forecast mean daily temperature data at 1st and 2nd Perak BMKG Station in Surabaya, this study used the univariate method, ARIMA model and multivariate method, VARIMA model with outlier detection. The best ARIMA model was selected using in-sample criteria, i.e. AIC and BIC. While for VAR model, the minimum information criterion namely AICc value was considered. The RMSE values of several forecasting horizons of out-sample data showed that the overall best model for mean daily temperature at 1st and 2nd Perak Station was the multivariate model, i.e. VARX (10,1) with four outliers incorporated in the model, indicated that it was necessary to consider the temperature from the nearest stations to improve the forecasting performance. This study recommends performing the overall best model only for short term forecasting, i.e. two weeks at maximum. By using the one week-step ahead and one day-step ahead forecasting scheme, the forecasting performance is significantly improved compared to default the k-step ahead forecasting scheme.

Machine Learning in Assessing the Performance of Hydrological Models

Machine learning has been employed successfully as a tool virtually in every scientific and technological field. In hydrology, machine learning models first appeared as simple feed-forward networks that were used for short-term forecasting, and have evolved into complex models that can take into account even the static features of catchments, imitating the hydrological experience. Recent studies have found machine learning models to be robust and efficient, frequently outperforming the standard hydrological models (both conceptual and physically based). However, and despite some recent efforts, the results of the machine learning models require significant effort to interpret and derive inferences. Furthermore, all successful applications of machine learning in hydrology are based on networks of fairly complex topology that require significant computational power and CPU time to train. For these reasons, the value of the standard hydrological models remains indisputable. In this study, we suggest employing machine learning models not as a substitute for hydrological models, but as an independent tool to assess their performance. We argue that this approach can help to unveil the anomalies in catchment data that do not fit in the employed hydrological model structure or configuration, and to deal with them without compromising the understanding of the underlying physical processes.